990 9'ournal of Neurology,
Motor recovery
Neurosurgery, and Psychiatry 1992;55:990-996
after acute ischaemic stroke:
a
metabolic study V Di Piero, F M Chollet, P MacCarthy, G L Lenzi, R S J Frackowiak
MRC Cyclotron Unit, Hanmersmith Hospital, London, UK V Di Piero* R S J Frackowiak F M Chollett Cattedra di Neuropatologia e Psicopatologia,* Dipartimento di Scienze Neurologiche, Universita degli Studi "La Sapienza", Rome, Italy G L Lenzi Service de Neurologie,t HBpital Purpan, Toulouse, France
Department of Radiology, Hammersmith Hospital, London, UK P MacCarthy Correspondence to: Professor Frackowiak, MRC Cyclotron Unit, Hammersmith Hospital, 150 Ducane Road, Lond6n W12 OHS, UK Received 13 February 1991 and in final revised form 31 October 1991. Accepted 14 December 1991
Abstract The metabolic changes occurring after ischaemic stroke were measured to investigate the functional anatomy of clinical motor recovery. Positron emission tomography (PET) and the steady-state 15O technique was used to compare resting relative metabolic distributions at the onset of functional deficit with those following recovery. Ten patients were studied with repeat scans. Motor recovery was associated in some patients with an increase of relative oxygen metabolism in anatomical structures normally involved in motor function in the affected hemisphere, particularly in the cortical motor areas. In those patients without such metabolic changes in the cortex of the diseased hemisphere, relative increases in cortical metabolism in the contralateral hemisphere were associated with better motor recovery than in patients with no relative cortical metabolic increase in either hemisphere. There was no correlation between the degree of improvement in motor function and the severity of motor deficit at onset, the size and site of the lesion and the metabolic changes in the infarcted zone. No particular pattern of global metabolic changes was observed after recovery. Thus different relative patterns of metabolic recovery were seen in patients with different lesions and evidence was found for the participation of contralateral structures in the recovery process in some patients.
basis for such recovery. The investigation of such aspects of human functional recovery has been mainly clinical and pathological with attention directed principally at the ischaemic lesion. The results have been contradictory.25 Cerebral blood flow (CBF) has been the principal functional parameter investigated with the development of functional neuroimaging techniques. However, in the acute phase of stroke there is profound mismatch between cerebral metabolism and blood flow with various degrees of relative or absolute hyperaemia, -' which means that CBF does not reflect cerebral function in the early phase of injury. The mismatch is responsible for the low prognostic value of CBF unless the measurement is made within a very few hours of onset of the acute ischaemia.9"-1 Positron emission tomography (PET) can be used to measure human brain metabolism in vivo thus providing more direct evidence of cerebral function in the acute phase of stroke. The aim of our study was to evaluate the metabolic changes occurring after ischaemic stroke in the acute and recovered phase and to investigate correlations with clinical recovery of motor function.
Materials and methods We studied 5 male and 5 female patients, mean (SD) age 54-3 (8 7) years, presenting with motor dysfunction due to acute ischaemic infarction in a cerebral hemisphere. Comatose patients or those suffering from other neurological conditions were excluded. The study had the approval of the Hammersmith Hospi(7 Neurol Neurosurg Psychiatry 1992;55:990-996) tal ethics committee and the Administration of Radioactive Substances Advisory Committee of the Department of Health, UK. Written A demonstrably effective therapeutic approach informed consent was obtained in each case to acute cerebral ischaemia has still to be from the patient or the nearest relative. All but defined. This is partly due to an incomplete one patient (case 8-table 1) had no history of understanding of the mechanisms leading to cerebrovascular disease. Clinical evaluation ischaemic cell death and partly to an even included a full neurological examination, an greater ignorance about the mechanisms of assessment of motor function using the motricfunctional recovery following cerebral infarc- ity index (MI) of Demeurisse et al, 2 of general tion. Research in the area of functional rehabil- function using the Barthel scale,'3 a Mini itation is doubly difficult because of the Mental State test'4 and a formal neuropsychomultiple variables that affect outcome and the logical evaluation. The MI is a published scale difficulty of selecting homogeneous patient based upon the MRC grading of muscle populations with respect to mechanism and strength, especially designed to evaluate extent of infarction.1 Methods of selecting "patients' progress in motor recovery ... and groups of patients that might benefit from correlations with other clinical data".'2 The intensive rehabilitation treatment and what scale reflects motor performance using three forms, if any, such treatment should take movements in each limb with appropriate remain to be defined. One step would be the weighting to give a 0-100 scale, with 100 establishment of the functional anatomical representing normality. Scores from 0 to 32
991
Motor recovery after acute ischaemic stroke: a metabolic study
Table 1 Clinical presentation Days from N Age Sex stroke onset 1
62 m
3
CT orMRI scan lesion
Time interval to Size second scan
1 Facio-brachial paresis 1 arm hypoesthesia
r fronto-temporo-
1
202
r Internal capsule (MR)
s
150
r Internal capsule
s
121
s
60
s
92
1
Clinical presentation
parietal
3 53 f
1
1 Hemiparesis 1 Hypoesthesis and constructional apraxia Hemiparesis
4 58 f
5
1 arm paresis
56 f
17
r Hemiparesis
r Internal capsule and thalamus (MRI) 1 Corona radiata r White matter lucencies (MRI)
6 34 f
5
r Hemiparesis, apraxia and
Superior fronto-parietal
transcortical aphasia lack of initiative and concentration difficulties 1 Hemiparesis
r Anterior white
matter I Nucleus caudate r Internal capsule r Fronto temporal
rThalamic
s
2
5
7
62 f
53 m
3
7
64 m
3
1 Hemiparesis
9 49 m
2
1 Ataxic hemiparesis slurred speech r Hemiparesis (arm>>leg) r hypoesthesia aphasia
8
10 52 m
3
I Fronto-temporoparietal
Residual deficit Minor difficulties fine finger movements and hand hypoesthesia Increased tone and hypoesthesia
Motricity Index I
II
88
, 100
92
100
Other clinical data
Doppler normal previous myocardial infarct High blood pressure
doppler I int carotid stenosis Angina
Mild weakness 1 finger abduction Increased tone and paresthesiae Hemiparesis (arm>leg)
71
100
72
86
High blood pressure
0
35
High blood pressure rheumatic fever
86
Emiparesis (leg>>arm) mild perseveratio
26
71
Angio: occ I supramarginal a lupus anticoagulant
s
62
Absent
95
100
s s s
63
Absent
87
100
60
Absent
86
100
High blood pressure Peripheral vascular disease High blood pressure 1982 r CVA with fill recovery previous myocardial Infarction Angiography: normal
55
r Mild hemiparesis
44
90
High blood pressure
1
r Hypoesthesia motor aphasia
indicate a severe plegia, from 33-64 a moderate paresis and 65-99 mild disability. Improvement in motor function was assessed by the raw score increment obtained at follow up examination relative to the initial assessment during the acute phase of the illness. Ratings were performed by one assessor throughout to obviate added variance resulting from interrater variability. All patients had CT or MRI scanning on admission and, if necessary, this was repeated following recovery. Delayed CT or MRI scans were used to delineate the extent of the infarction. All images were reoriented so that the transverse slices lay parallel to the intercommissural plane. All the patients received standard medical treatment for their strokepatients 4, 5, 7, 8 and 10 received antihypertensive medication, none received Ca channel blockers, anticonvulsants, sedatives or anti-psychotics. Clinical data are summarised in table 1. PET studies PET was performed at a mean (SD) 4 9 (4 6) days from symptom onset and was repeated in the chronic phase [95- 1 (48 7) days], following appreciable recovery of function. The "5O inhalation method was used to measure cerebral blood flow (CBF), cerebral blood volume (CBV) and oxygen extraction (OER), from which the rate of oxygen consumption (CMRO2) was calculated.'5 16 Studies were carried out with an ECAT 931/8/12 (CTI, Knoxville) and performed at rest, in a quiet and dimly lit room with the patients' eyes closed and ears unplugged. During each study, head position was maintained by an individually moulded thermoplastic head holder. A
Atrial fibrillation
short 21 g arterial line was placed into the radial artery at the onset of each study. The intrinsic spatial resolution of tWe scanner was 5-5 x 5.5 x 70 mm full-width at half maximum (FWHM), allowing the simultaneous collection of 15 contiguous transaxial planes.'7 Reconstruction, attenuation correction (by measurement) and filtering resulted in images with a resolution of 8-5 x 8 5 x 7-0 mm (FWHM). The set of data was then expanded by linear interpolation in the axial dimension to produce 43 transaxial slices with near cubic voxels of 2.05 x 2.05 x 2.5 mm allowing three dimensional inspection of scans visually. The scans were analysed on a computer (SUN 3/60) with image analysis software (ANALYZE, BRU, Mayo Clinic), which allowed them to be displayed relative to the intercommisural line (AC-PC line) and to be scaled for precise anatomical localisation by standard stereotactic coordinates based on the Talairach atlas."8-9"The procedure of reorienting PET scans was developed for normal studies but has been used successfully in studies of patients with cerebral infarcts. Data analysis In the acute phase of stroke a mismatch may occur between CMRO2 and CBF which manifests as a change in OER.7 For this reason, we analysed CMRO2 data which give a more direct reflection of tissue metabolism than CBF. Variability in the absolute levels of global CMRO2 between patients and between successive measurements in individuals averaged 9-4 (4-8)%. Such variability was present in both the lesioned and unlesioned hemispheres and is in accordance with previous data.2021 Even such small interscan variability may mask
992
Di Piero, Chollet, MacCarthy, Lenzi, Frackowiak
significant regional (r)CMRO2 changes. To circumvent this problem, and because we were primarily interested in variations in the regional profile of metabolism between the acute and recovered phases, we normalised the rCMRO2 data of the second scan of each patient to the average CMRO2 of the patient's first scan. Three features were analysed in the PET scans: 1) relative rCMRO2 changes in the anatomical area of the ischaemic lesion; 2) global CMRO2 changes occurring between the first and second scans in the group of patients; 3) relative rCMRO2 changes in individual patients. 1) Changes in the area of the lesion: To study the variations occurring in the area of the lesion, each pair of studies was aligned parallel to the AC-PC line."8 Irregular regions of interest (ROI) were drawn interactively, using the anatomical boundaries of the lesion shown on the similarly reoriented CT or MRI scan, onto the PET scan planes. The same ROIs were then automatically placed on the homologous slices of the follow up study. The average of the rCMRO2 value for the whole lesion so defined was compared with the average normalised rCMRO2 from the same region on the second scan for each patient. 2) Changes as a group: We used a descriptive and exploratory statistical approach to look for any general metabolic pattern associated with motor recovery. All the studies, irrespective of the site of the lesion were considered. The scans were flipped about the midsagittal plane such that all the ischaemic lesions were analysed as though left sided. All the scans were also resized by linear scaling into standardised brain dimensions. Data were smoothed by a square low-pass filter of size 9 pixels to improve the signal to noise and reduce variance due to differences in gyral anatomy between individuals. The presence of significant relative increases in rCMRO2 was assessed by constructing average pixel-by-pixel maps of the normalised rCMRO2 and associated variances for the first and the second scans and then comparing them by a t test. In this way, areas of significant change were shown as an image of t values thresholded for descriptive purposes at two levels: p < 0-01 and p < 0001.The anatomical localisation of significantly changed areas was performed by simple cross reference to coordinates in the stereotaxic atlas ofTalairach and Tournoux. 9 3) Individual changes: To assess individual changes in relative rCMRO2 we considered each pair of studies matched anatomically by reference to the AC-PC line. Scans were smoothed, again to increase signal to noise, though less severely because of the absence of interindividual anatomical variation. Subtraction images were generated by taking the CMRO2 values of the first scan from the second, pixel-by-pixel. Only regions with increases spread over 4 contiguous pixels and anatomically associated with the motor system were considered further, that is, the primary motorsensory area, the premotor area, the supplementary motor area, the basal ganglia and the thalamus. Increases in normalised
rCMRO2 were quantified and expressed as percentage changes. The localisation of areas with an rCMRO2 increase was performed as above. Data from the cerebellar regions were not available from all patients and therefore are not analysed further in this paper. In an attempt to describe with one variable both the extent and magnitude of relative rCMRO2 changes, we defined an index (oxygen metabolism index: OMI) from the product of the per cent normalised rCMRO2 increase in the region (calculated as the ratio of the average rCMRO, between the second and the first scans) and the size of the contiguous area over which change was demonstrated (in pixels). Statistical analysis Changes in rCMRO2 and OMI values and clinical scores were analysed by intergroup comparisons using Student's t test for paired and unpaired observations, linear regressions and Spearman rank correlations. Bonferroni corrections were used to adjust the threshold of significance for multiple comparisons. Results At the second examination, all patients demonstrated some recovery of motor function. The mean (SD) index of motor recovery (RI = MI at follow up - MI at entry), was 22 (15) (+33%), ranging from +5 to +46. The MI at entry was positively correlated to the MI at follow up (r = 0 91; p < 0 000 1), but there was no correlation between the RI and both the MI at entry or at follow up scanning. Three patients (1, 6 and 10) had large cortical lesions on CT or MRI, while the remaining seven had small infarcts with a maximum diameter less than 2-5 cm. There were no statistically significant differences between the MI at entry or follow up, or of the RI between patients with small and large lesions. In the lesion, a relative rCMRO2 increase was observed in all but three patients (1, 4 and 5). On average, the rCMRO2 increase was 9 (17) %. There was no correlation between the rCMRO2 changes at the site of lesion and the RI (fig 1). Statistical analysis failed to reveal a single pattern of rCMRO2 change for the 50
40 U
RI
30 20
10
.~~ 0
2 1 CMR02 Increase (2nd/lst)
Figure 1 Variations of CMROQ in the area of the anatomical lesion, as a ratio between the second and the first scan (abscissa) and the degree of improvement in motor function (RI).
993
Motor recovery after acute ischaemic stroke: a metabolic study
Table 2 Image subtraction analysis: CMRO2 changes after motor recovery evaluated as increases in the oxygen metabolism index (OMI) in cerebral regions normally associated with the generation of movements Case number CT or MRI scan
RI
Affected hemisphere regions
IIII
n° pixels OMI
I
r Fronto-temporo-parietal
12
-
-
-
-
2 3
r Internal capsule r Internal capsule
8 29
basal ganglia 1 motorsensory area supplementary motor area basal ganglia thalamus -
1-63 1-28 1-16 1-26 1-37 -
20 80 16 16 16 -
32-56 102-40 18-56 17-26 21-92 -
1 motorsensory area basal ganglia thalamus supplementary motor area 1 motorsensory area thalamus -
1-23 1-30 1-48 1-75 1-47 1-41
16 16 32 112 112 16
19-68 20-80 47-36 195.84 164-48 22-56
4 5 6 7
8 9
10
14
r Internal capsule and thalamus (MRI) I corona radiata r white matter "lucencies" (MRI) 1 fronto-parietal
35
45
r anterior white matter I nucleus caudatus r Internal capsule 1 Fronto-temporal r thalamus
13
1 motorsensory area
1-57
16
25-12
14
1 motorsensory area
1-46
64
93-24
fronto-temporo-parietale sn
46
1 motor sensory area supplementary motor area
1-31 1-37
64 16
83-68 21-92
5
-
-
group of patic!nts as a whole and for patients grouped accor-ding to the size of the ischaemic lesion. The results of individual analyses of changes in motor ass;ociate4 areas are reported in table 2. The magnitude and extent of metabolic improveiment did not correlate with MI scores at entr3y or at follow up. The change in OMI did not cliffer between patients with large and small is chaemic lesions in either the affected or nnaffected cerebral hemispheres, neither regionLally nor globally. A correlatio)n between the RI and the OMI in cerebral r egions associated with motor control was c)bserved in the affected hemisphere (r = 0-75; p < 0-01) (fig 2). No correlation waLs found between motor recovery and metaboliic increases in the unaffected hemisphere or in the motor areas of both sides of the brain pooled. There was no significant difference in tthe OMI between the two cerebral hemisphe res. We also anialysed the different individual patterns of rC,MRO2 change (table 3). In the cortex one pa tient showed an increase in the
50
40
RI
p
Motor recovery
Neurosurgery, and Psychiatry 1992;55:990-996
after acute ischaemic stroke:
a
metabolic study V Di Piero, F M Chollet, P MacCarthy, G L Lenzi, R S J Frackowiak
MRC Cyclotron Unit, Hanmersmith Hospital, London, UK V Di Piero* R S J Frackowiak F M Chollett Cattedra di Neuropatologia e Psicopatologia,* Dipartimento di Scienze Neurologiche, Universita degli Studi "La Sapienza", Rome, Italy G L Lenzi Service de Neurologie,t HBpital Purpan, Toulouse, France
Department of Radiology, Hammersmith Hospital, London, UK P MacCarthy Correspondence to: Professor Frackowiak, MRC Cyclotron Unit, Hammersmith Hospital, 150 Ducane Road, Lond6n W12 OHS, UK Received 13 February 1991 and in final revised form 31 October 1991. Accepted 14 December 1991
Abstract The metabolic changes occurring after ischaemic stroke were measured to investigate the functional anatomy of clinical motor recovery. Positron emission tomography (PET) and the steady-state 15O technique was used to compare resting relative metabolic distributions at the onset of functional deficit with those following recovery. Ten patients were studied with repeat scans. Motor recovery was associated in some patients with an increase of relative oxygen metabolism in anatomical structures normally involved in motor function in the affected hemisphere, particularly in the cortical motor areas. In those patients without such metabolic changes in the cortex of the diseased hemisphere, relative increases in cortical metabolism in the contralateral hemisphere were associated with better motor recovery than in patients with no relative cortical metabolic increase in either hemisphere. There was no correlation between the degree of improvement in motor function and the severity of motor deficit at onset, the size and site of the lesion and the metabolic changes in the infarcted zone. No particular pattern of global metabolic changes was observed after recovery. Thus different relative patterns of metabolic recovery were seen in patients with different lesions and evidence was found for the participation of contralateral structures in the recovery process in some patients.
basis for such recovery. The investigation of such aspects of human functional recovery has been mainly clinical and pathological with attention directed principally at the ischaemic lesion. The results have been contradictory.25 Cerebral blood flow (CBF) has been the principal functional parameter investigated with the development of functional neuroimaging techniques. However, in the acute phase of stroke there is profound mismatch between cerebral metabolism and blood flow with various degrees of relative or absolute hyperaemia, -' which means that CBF does not reflect cerebral function in the early phase of injury. The mismatch is responsible for the low prognostic value of CBF unless the measurement is made within a very few hours of onset of the acute ischaemia.9"-1 Positron emission tomography (PET) can be used to measure human brain metabolism in vivo thus providing more direct evidence of cerebral function in the acute phase of stroke. The aim of our study was to evaluate the metabolic changes occurring after ischaemic stroke in the acute and recovered phase and to investigate correlations with clinical recovery of motor function.
Materials and methods We studied 5 male and 5 female patients, mean (SD) age 54-3 (8 7) years, presenting with motor dysfunction due to acute ischaemic infarction in a cerebral hemisphere. Comatose patients or those suffering from other neurological conditions were excluded. The study had the approval of the Hammersmith Hospi(7 Neurol Neurosurg Psychiatry 1992;55:990-996) tal ethics committee and the Administration of Radioactive Substances Advisory Committee of the Department of Health, UK. Written A demonstrably effective therapeutic approach informed consent was obtained in each case to acute cerebral ischaemia has still to be from the patient or the nearest relative. All but defined. This is partly due to an incomplete one patient (case 8-table 1) had no history of understanding of the mechanisms leading to cerebrovascular disease. Clinical evaluation ischaemic cell death and partly to an even included a full neurological examination, an greater ignorance about the mechanisms of assessment of motor function using the motricfunctional recovery following cerebral infarc- ity index (MI) of Demeurisse et al, 2 of general tion. Research in the area of functional rehabil- function using the Barthel scale,'3 a Mini itation is doubly difficult because of the Mental State test'4 and a formal neuropsychomultiple variables that affect outcome and the logical evaluation. The MI is a published scale difficulty of selecting homogeneous patient based upon the MRC grading of muscle populations with respect to mechanism and strength, especially designed to evaluate extent of infarction.1 Methods of selecting "patients' progress in motor recovery ... and groups of patients that might benefit from correlations with other clinical data".'2 The intensive rehabilitation treatment and what scale reflects motor performance using three forms, if any, such treatment should take movements in each limb with appropriate remain to be defined. One step would be the weighting to give a 0-100 scale, with 100 establishment of the functional anatomical representing normality. Scores from 0 to 32
991
Motor recovery after acute ischaemic stroke: a metabolic study
Table 1 Clinical presentation Days from N Age Sex stroke onset 1
62 m
3
CT orMRI scan lesion
Time interval to Size second scan
1 Facio-brachial paresis 1 arm hypoesthesia
r fronto-temporo-
1
202
r Internal capsule (MR)
s
150
r Internal capsule
s
121
s
60
s
92
1
Clinical presentation
parietal
3 53 f
1
1 Hemiparesis 1 Hypoesthesis and constructional apraxia Hemiparesis
4 58 f
5
1 arm paresis
56 f
17
r Hemiparesis
r Internal capsule and thalamus (MRI) 1 Corona radiata r White matter lucencies (MRI)
6 34 f
5
r Hemiparesis, apraxia and
Superior fronto-parietal
transcortical aphasia lack of initiative and concentration difficulties 1 Hemiparesis
r Anterior white
matter I Nucleus caudate r Internal capsule r Fronto temporal
rThalamic
s
2
5
7
62 f
53 m
3
7
64 m
3
1 Hemiparesis
9 49 m
2
1 Ataxic hemiparesis slurred speech r Hemiparesis (arm>>leg) r hypoesthesia aphasia
8
10 52 m
3
I Fronto-temporoparietal
Residual deficit Minor difficulties fine finger movements and hand hypoesthesia Increased tone and hypoesthesia
Motricity Index I
II
88
, 100
92
100
Other clinical data
Doppler normal previous myocardial infarct High blood pressure
doppler I int carotid stenosis Angina
Mild weakness 1 finger abduction Increased tone and paresthesiae Hemiparesis (arm>leg)
71
100
72
86
High blood pressure
0
35
High blood pressure rheumatic fever
86
Emiparesis (leg>>arm) mild perseveratio
26
71
Angio: occ I supramarginal a lupus anticoagulant
s
62
Absent
95
100
s s s
63
Absent
87
100
60
Absent
86
100
High blood pressure Peripheral vascular disease High blood pressure 1982 r CVA with fill recovery previous myocardial Infarction Angiography: normal
55
r Mild hemiparesis
44
90
High blood pressure
1
r Hypoesthesia motor aphasia
indicate a severe plegia, from 33-64 a moderate paresis and 65-99 mild disability. Improvement in motor function was assessed by the raw score increment obtained at follow up examination relative to the initial assessment during the acute phase of the illness. Ratings were performed by one assessor throughout to obviate added variance resulting from interrater variability. All patients had CT or MRI scanning on admission and, if necessary, this was repeated following recovery. Delayed CT or MRI scans were used to delineate the extent of the infarction. All images were reoriented so that the transverse slices lay parallel to the intercommissural plane. All the patients received standard medical treatment for their strokepatients 4, 5, 7, 8 and 10 received antihypertensive medication, none received Ca channel blockers, anticonvulsants, sedatives or anti-psychotics. Clinical data are summarised in table 1. PET studies PET was performed at a mean (SD) 4 9 (4 6) days from symptom onset and was repeated in the chronic phase [95- 1 (48 7) days], following appreciable recovery of function. The "5O inhalation method was used to measure cerebral blood flow (CBF), cerebral blood volume (CBV) and oxygen extraction (OER), from which the rate of oxygen consumption (CMRO2) was calculated.'5 16 Studies were carried out with an ECAT 931/8/12 (CTI, Knoxville) and performed at rest, in a quiet and dimly lit room with the patients' eyes closed and ears unplugged. During each study, head position was maintained by an individually moulded thermoplastic head holder. A
Atrial fibrillation
short 21 g arterial line was placed into the radial artery at the onset of each study. The intrinsic spatial resolution of tWe scanner was 5-5 x 5.5 x 70 mm full-width at half maximum (FWHM), allowing the simultaneous collection of 15 contiguous transaxial planes.'7 Reconstruction, attenuation correction (by measurement) and filtering resulted in images with a resolution of 8-5 x 8 5 x 7-0 mm (FWHM). The set of data was then expanded by linear interpolation in the axial dimension to produce 43 transaxial slices with near cubic voxels of 2.05 x 2.05 x 2.5 mm allowing three dimensional inspection of scans visually. The scans were analysed on a computer (SUN 3/60) with image analysis software (ANALYZE, BRU, Mayo Clinic), which allowed them to be displayed relative to the intercommisural line (AC-PC line) and to be scaled for precise anatomical localisation by standard stereotactic coordinates based on the Talairach atlas."8-9"The procedure of reorienting PET scans was developed for normal studies but has been used successfully in studies of patients with cerebral infarcts. Data analysis In the acute phase of stroke a mismatch may occur between CMRO2 and CBF which manifests as a change in OER.7 For this reason, we analysed CMRO2 data which give a more direct reflection of tissue metabolism than CBF. Variability in the absolute levels of global CMRO2 between patients and between successive measurements in individuals averaged 9-4 (4-8)%. Such variability was present in both the lesioned and unlesioned hemispheres and is in accordance with previous data.2021 Even such small interscan variability may mask
992
Di Piero, Chollet, MacCarthy, Lenzi, Frackowiak
significant regional (r)CMRO2 changes. To circumvent this problem, and because we were primarily interested in variations in the regional profile of metabolism between the acute and recovered phases, we normalised the rCMRO2 data of the second scan of each patient to the average CMRO2 of the patient's first scan. Three features were analysed in the PET scans: 1) relative rCMRO2 changes in the anatomical area of the ischaemic lesion; 2) global CMRO2 changes occurring between the first and second scans in the group of patients; 3) relative rCMRO2 changes in individual patients. 1) Changes in the area of the lesion: To study the variations occurring in the area of the lesion, each pair of studies was aligned parallel to the AC-PC line."8 Irregular regions of interest (ROI) were drawn interactively, using the anatomical boundaries of the lesion shown on the similarly reoriented CT or MRI scan, onto the PET scan planes. The same ROIs were then automatically placed on the homologous slices of the follow up study. The average of the rCMRO2 value for the whole lesion so defined was compared with the average normalised rCMRO2 from the same region on the second scan for each patient. 2) Changes as a group: We used a descriptive and exploratory statistical approach to look for any general metabolic pattern associated with motor recovery. All the studies, irrespective of the site of the lesion were considered. The scans were flipped about the midsagittal plane such that all the ischaemic lesions were analysed as though left sided. All the scans were also resized by linear scaling into standardised brain dimensions. Data were smoothed by a square low-pass filter of size 9 pixels to improve the signal to noise and reduce variance due to differences in gyral anatomy between individuals. The presence of significant relative increases in rCMRO2 was assessed by constructing average pixel-by-pixel maps of the normalised rCMRO2 and associated variances for the first and the second scans and then comparing them by a t test. In this way, areas of significant change were shown as an image of t values thresholded for descriptive purposes at two levels: p < 0-01 and p < 0001.The anatomical localisation of significantly changed areas was performed by simple cross reference to coordinates in the stereotaxic atlas ofTalairach and Tournoux. 9 3) Individual changes: To assess individual changes in relative rCMRO2 we considered each pair of studies matched anatomically by reference to the AC-PC line. Scans were smoothed, again to increase signal to noise, though less severely because of the absence of interindividual anatomical variation. Subtraction images were generated by taking the CMRO2 values of the first scan from the second, pixel-by-pixel. Only regions with increases spread over 4 contiguous pixels and anatomically associated with the motor system were considered further, that is, the primary motorsensory area, the premotor area, the supplementary motor area, the basal ganglia and the thalamus. Increases in normalised
rCMRO2 were quantified and expressed as percentage changes. The localisation of areas with an rCMRO2 increase was performed as above. Data from the cerebellar regions were not available from all patients and therefore are not analysed further in this paper. In an attempt to describe with one variable both the extent and magnitude of relative rCMRO2 changes, we defined an index (oxygen metabolism index: OMI) from the product of the per cent normalised rCMRO2 increase in the region (calculated as the ratio of the average rCMRO, between the second and the first scans) and the size of the contiguous area over which change was demonstrated (in pixels). Statistical analysis Changes in rCMRO2 and OMI values and clinical scores were analysed by intergroup comparisons using Student's t test for paired and unpaired observations, linear regressions and Spearman rank correlations. Bonferroni corrections were used to adjust the threshold of significance for multiple comparisons. Results At the second examination, all patients demonstrated some recovery of motor function. The mean (SD) index of motor recovery (RI = MI at follow up - MI at entry), was 22 (15) (+33%), ranging from +5 to +46. The MI at entry was positively correlated to the MI at follow up (r = 0 91; p < 0 000 1), but there was no correlation between the RI and both the MI at entry or at follow up scanning. Three patients (1, 6 and 10) had large cortical lesions on CT or MRI, while the remaining seven had small infarcts with a maximum diameter less than 2-5 cm. There were no statistically significant differences between the MI at entry or follow up, or of the RI between patients with small and large lesions. In the lesion, a relative rCMRO2 increase was observed in all but three patients (1, 4 and 5). On average, the rCMRO2 increase was 9 (17) %. There was no correlation between the rCMRO2 changes at the site of lesion and the RI (fig 1). Statistical analysis failed to reveal a single pattern of rCMRO2 change for the 50
40 U
RI
30 20
10
.~~ 0
2 1 CMR02 Increase (2nd/lst)
Figure 1 Variations of CMROQ in the area of the anatomical lesion, as a ratio between the second and the first scan (abscissa) and the degree of improvement in motor function (RI).
993
Motor recovery after acute ischaemic stroke: a metabolic study
Table 2 Image subtraction analysis: CMRO2 changes after motor recovery evaluated as increases in the oxygen metabolism index (OMI) in cerebral regions normally associated with the generation of movements Case number CT or MRI scan
RI
Affected hemisphere regions
IIII
n° pixels OMI
I
r Fronto-temporo-parietal
12
-
-
-
-
2 3
r Internal capsule r Internal capsule
8 29
basal ganglia 1 motorsensory area supplementary motor area basal ganglia thalamus -
1-63 1-28 1-16 1-26 1-37 -
20 80 16 16 16 -
32-56 102-40 18-56 17-26 21-92 -
1 motorsensory area basal ganglia thalamus supplementary motor area 1 motorsensory area thalamus -
1-23 1-30 1-48 1-75 1-47 1-41
16 16 32 112 112 16
19-68 20-80 47-36 195.84 164-48 22-56
4 5 6 7
8 9
10
14
r Internal capsule and thalamus (MRI) I corona radiata r white matter "lucencies" (MRI) 1 fronto-parietal
35
45
r anterior white matter I nucleus caudatus r Internal capsule 1 Fronto-temporal r thalamus
13
1 motorsensory area
1-57
16
25-12
14
1 motorsensory area
1-46
64
93-24
fronto-temporo-parietale sn
46
1 motor sensory area supplementary motor area
1-31 1-37
64 16
83-68 21-92
5
-
-
group of patic!nts as a whole and for patients grouped accor-ding to the size of the ischaemic lesion. The results of individual analyses of changes in motor ass;ociate4 areas are reported in table 2. The magnitude and extent of metabolic improveiment did not correlate with MI scores at entr3y or at follow up. The change in OMI did not cliffer between patients with large and small is chaemic lesions in either the affected or nnaffected cerebral hemispheres, neither regionLally nor globally. A correlatio)n between the RI and the OMI in cerebral r egions associated with motor control was c)bserved in the affected hemisphere (r = 0-75; p < 0-01) (fig 2). No correlation waLs found between motor recovery and metaboliic increases in the unaffected hemisphere or in the motor areas of both sides of the brain pooled. There was no significant difference in tthe OMI between the two cerebral hemisphe res. We also anialysed the different individual patterns of rC,MRO2 change (table 3). In the cortex one pa tient showed an increase in the
50
40
RI
p
